This research is concerned with developing and testing a theory of human cognition. That theory sees human cognition as operating according to a set of production rules. Each production rule specifies that a mental act should take place when a particular type of mental state occurs. The research will continue the development of the PUPS computer simulation program, begun under Anderson's prior NSF support, which enables one to trace out the interactions of these production rules. An effort will be made to make the PUPS simulation available to other interested researchers in the cognitive science community. Part of the effort will be devoted to the writing of a research monograph that will describe the support for Anderson's production system theory. Empirical research will be devoted to analyzing the distinction between procedural and declarative knowledge which is central to the theory. Procedural knowledge is knowledge about how to perform various cognitive skills and is encoded as production rules. Declarative knowledge is factual knowledge about the world which can be used by the production rules. The research will try to establish that these are separate types of knowledge, by showing that there is little transfer between learning procedural and learning declarative knowledge, that declarative knowledge is more readily interfered with by performing a concurrent task which involves similar information, and that people have more conscious control over expression of declarative knowledge. The research will also compare the basic parameters associated with acquisition and retention of the two types of knowledge. The theory is of general applied significance, because it is concerned with the processes by which complex cognitive skills (like doing a proof in logic) are acquired. In particular, in examining the relationship between declarative and procedural knowledge, the research is laying the foundation for understanding how people make the transition from reading written instruction in a technical domain to solving problems fluently in that domain. Anderson is following up this applied challenge in other projects with other funding.
